Junxu Wang

Junxu Wang

Auto-generated profile, based on publicly available information

Junxu Wang

School of Electronics and Communication Engineering, Sun Yat-sen University, Shenzhen, China

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 July 2020 in IEEE Access
Verified
Journal Article
Open access
CDChuan Du
Chuan Du
LZLei Zhang
Lei Zhang
XSXiping Sun
Xiping Sun
JWJunxu Wang
Junxu Wang
JSJialian Sheng
Jialian Sheng
Millimeter-wave (MMW) radar hand gesture recognition technology is becoming important in many electronic device control applications. Currently, most existing approaches utilize the radical and micro-Doppler features from single-channel MMW radar, which ignores the different importance of the information contained in the micro-Doppler feature background or target areas. In this paper, we propose an algorithm for hand gesture recognition jointly using multi-channel signatures. The algorithm blends the information of both micro-Doppler features and instantaneous angles (azimuth and elevation) to accomplish hand gesture recognition performed with the convolutional neural network (CNN). To have a better features fusion and make CNN focus on the most important target signal regions and suppress the unnecessary noise areas, we apply the channel and spatial attention-based feature refinement modules. We also employ gesture movement mechanism-based data augmentation for more effective training to alleviate potential overfitting. Extensive experiments demonstrate the effectiveness and superiorities of the proposed algorithm. This method achieves a correct classification rate of 96.61%, approximately 5% higher than that of the single-channel-based recognition strategy as measured based on MMW radar datasets.

2 October 2023 in IEEE Geoscience and Remote Sensing Letters
Verified
Journal Article
Closed access
JWJunxu Wang
Junxu Wang
LZLei Zhang
Lei Zhang
ZMzhichao meng
zhichao meng
SWShaopeng Wei
Shaopeng Wei
LCLi-Li Chen
Li-Li Chen
Radio frequency interference (RFI) has become a serious problem for synthetic aperture radar (SAR) systems. In recent years, numerous methods have emerged for the suppression of radio frequency interference (RFI). These methods often rely on assumptions regarding the narrowband characteristics and low-rank nature of radar echoes or interference signals. However, these assumptions may not hold true when dealing with complex RFI. To mitigate complex interference in dual-polarization (dual-pol) synthetic aperture radar (SAR) systems, a proposed method is the utilization of a filtering algorithm based on sidelobe cancellation (SLC) technology. This method leverages the discrepancies in gain and coherence between observed targets and RFI signals in the co-polarization (co-pol) and cross-polarization (cross-pol) channels. It suppresses interference through SLC filtering on a range line by range line basis in the focused image domain, serving as a post-processing step. The method is capable of tackling multiple complex interferences optimally under the least mean square criterion, such as narrowband, wideband, and noise-modulated pulsed interferences. Both simulation and real data experiments show that the method can remove RFI artifacts and effectively recover the desired SAR images.

4 June 2025 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
ZMzhichao meng
zhichao meng
JWJunxu Wang
Junxu Wang
ZXZirui Xi
Zirui Xi
LZLei Zhang
Lei Zhang
JLJingyue Lu
Jingyue Lu
GWGuanyong Wang
Guanyong Wang
The existing Ground Cartesian Back-Projection (GCBP) algorithm is limited by its low effectiveness in spectral compression, which makes it unsuitable for processing high-squint and large-scale strip-map Synthetic Aperture Radar (SAR) data. To address this issue, we propose a novel algorithm called Ground Squint Fast Back Projection (GSFBP) for high-squint strip-map SAR imaging. First, we introduce a Ground Squint Coordinate (GSC) system that replaces the conventional Ground Cartesian Coordinate (GCC) system. The unique geometry of the GSC allows for precise rotation of the two-dimensional wavenumber spectrum without the need for auxiliary operations, significantly easing the constraints related to scene size.Moreover, the GSC framework facilitates seamless sub-image fusion through translation, eliminating the need for interpolation. Building on the original two-step spectral compression method, we have developed a GSC-specific relative spectrum inclination correction function to enhance spectral compression effectiveness. These innovations enable GSFBP to effectively manage large-scale scenes in high-squint SAR imaging. Experimental validation, using both simulated and real measured SAR data, confirms the superiority of the proposed algorithm.

1 January 2026 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
JWJunxu Wang
Junxu Wang
LZLei Zhang
Lei Zhang
SWShaopeng Wei
Shaopeng Wei
ZWZhenhua Wu
Zhenhua Wu
ZMzhichao meng
zhichao meng

21 May 2026 in IEEE Geoscience and Remote Sensing Letters
Verified
Journal Article
Closed access
SZSuijun Zhong
Suijun Zhong
QHQuan Huang
Quan Huang
JWJunxu Wang
Junxu Wang
LZLei Zhang
Lei Zhang
Deep unfolding networks based on the alternating direction method of multipliers (ADMM) have demonstrated significant potential in super-resolution inverse synthetic aperture radar (ISAR) imaging. However, in dynamic sensing scenarios characterized by varying sparse apertures and low signal-to-noise ratios, existing methods often suffer from limited adaptability and insufficient feature representation. To address these challenges, this letter proposes a complex-valued 2-D residual ADMM network (CV-2D-Res-ADMM-Net) for robust super-resolution ISAR imaging. Distinct from conventional implementations that rely on hand-crafted shrinkage operators, we design a learnable complex-valued 2-D residual module to implicitly learn the proximal mapping, thereby improving the reconstruction accuracy of complex structural details. Furthermore, a multisampling augmentation strategy is introduced during training to enhance the network’s generalization capability against diverse sparse sampling conditions. Experiments on measured data demonstrate that the proposed method achieves superior imaging performance and robustness compared to existing ADMM-based methods.